The US president election started a new chapter in the discussion on the role of data analytics in political marketing. Cambridge Analytica got a lot visibility when they conducted analytics for Trump’s campaign. At the same time there have been claims that Clinton’s campaign counted too much on analytics and micro-targeting, when bigger messages to larger audiences were missing. Are these conclusions contradictory or can different approaches explain the difference?
Targeting or even micro-targeting are not new in marketing and advertising. There are many ways to analyze consumer behavior to better to understand their preferences, needs and purchasing power. Typically consumers are divided into a few main segments or much higher numbers of micro-segments. We can also know the profile of each segment and target certain types of products, messages and pricing for them. Personalization means that we basically try to make segments of one person, but it is often too complex and expensive.
Profiling can be based on all kinds of available data, but in one of my earlier companies we divided it into three main categories: 1) demographics, 2) behavior, and 3) social network. One additional dimension is psychometrics that focuses more on understanding personality types.
Beyond Cambridge Analytica
Demographics mean, for example, age, gender, education, living area and family type. Behavior includes, what kinds of products you use and buy, which newspapers you read, what are your hobbies, what you like on Facebook and basically all data points that can be collected from your behavior. Social network data includes information with whom you are connected and how strongly, and do you have an influence on other people and who are the people that influence you. Nowadays more and more data is available to make richer profiling and analytics.
Often the main challenge of data analytics in marketing is not the analytics as such, but how to really utilize all profiling, segments and information in marketing activities. It is not easy to manage dozens or hundreds of different marketing offers and marketing messages. And it is not only a one time exercise, it also encapsulates monitoring changes in segments all the time, in order to see how they react to different campaigns and how messages and marketing actually works in different segments. The real bottleneck is typically having enough actionable information, processes to utilize the information, and continuous learning from real results.
Data analytics and targeting are not new in political campaigns. Parties and politicians have always tried to find people who are especially in their target group, where they can win new voters, and which messages work for different people. Technology and data analytics played an important part in Obama’s president campaigns too, and it was then said that technology was one key to his victory.
Cambridge Analytica was founded by its parent company, British SCL Group, especially for political marketing. It has worked especially with right-wing campaigns, like the Brexit campaign in the UK, as well as first with Ted Cruz and then Donald Trump in the US president election. Steve Bannon, a controversial person in Trump’s campaign and now White House chief strategist, has been on its board.
Not too many facts are known about Cambridge Analytica’s actual analytics. But it is believed to be based on the work of researcher Michal Kosinski in the Cambridge University to a certain degree, although he has nothing to do with the actual company. He worked with psychometric models, especially to measure five personality traits (openness to experience, conscientiousness, extraversion, agreeableness and neuroticism). He developed models for how a person’s profile and other characteristics can be predicted from his or her Facebook likes. One of their results was that it is possible to create very accurate profile for persons on less than 70 ‘like’ clicks.
But is this really something so new that it can explain the election results? It is really hard to say for sure, when we don’t know all details. But definitely Clinton’s campaign used a lot of analytics and targeting as well. Some political analysts have even commented she counted too much on tailored messages to women, other messages to African Americans and third ones to immigrants, but the big uniting messages were missing, when Trump had some very big catchy messages.
One answer can anyway be that the difference lies in how analytics results were used and how the feedback loop worked. Based on available information it looks like Cambridge Analytica and Trump’s campaign really focused to tailor messages for different target groups, not only assuming that certain segments like to hear something, but measure how target groups reacted to the messages, and also develop campaign’s main messages based on reactions in different groups.
They didn’t only use it to promote their own candidate in their potential groups, but e.g. sent negative messages to Clinton’s potential supporters and tried to get them to be passive. They didn’t only focus on voters’ preferences and interests, but also on their fears. So, we can say they were ready to use analytics in a very brutal way.
Many can claim that Trump’s messages were often contradictory and not very consistent. There are also claims that this is actually planned based on analytics; they had enough relevant points for different target groups. People often notice only messages that are relevant to them or fit in their predefined worldview. I talked once with the editor-and-chief of a newspaper. He asked me if I have noticed they had added the “bible verse of the day” to their newspaper. I hadn’t noticed that. He replied “I didn’t expect you to, it is a small box on a page, which truly dedicated Christians notice, love, and continue reading our paper. The rest don’t even notice it.”
We’ll probably see much more research and analysis from these elections and what the role of data and targeting really was. For me this discussion has especially raised three aspects that probably got too little attention in analytics:
- It is important to utilize analytics on messaging and content of campaigns, not only focus to know target groups, but measure how they like and react to different main and micro messages, and not make assumptions.
- People’s behavior is not driven only based on positive things, like what they target and like, but also what they hate and fear can be powerful tools in marketing - yet this raises ethical questions.
- Contradictory messages can work, when many only hear messages that are relevant and resonate with them.
Data analytics has become a fundamental part of all marketing. At the same time one campaign and success with one tactic doesn’t guarantee a long time sustainable position that is needed in brand building. It will be an on-going race to find new and better solutions all the time. At the same time it is also a race with privacy, ethics, and ownership of personal data.